UPTEC E 17 007
Examensarbete 30 hp Juni 2017
High Accuracy Speed and Angular Position Detection by Multiple Sensing
Nikola Stojanovic
Teknisk- naturvetenskaplig fakultet UTH-enheten
Besöksadress:
Ångströmlaboratoriet Lägerhyddsvägen 1 Hus 4, Plan 0 Postadress:
Box 536 751 21 Uppsala Telefon:
018 – 471 30 03 Telefax:
018 – 471 30 00 Hemsida:
http://www.teknat.uu.se/student
Abstract
High Accuracy Speed and Angular Position Detection by Multiple Sensing
Nikola Stojanovic
The information from numerous sensors placed in modern vehicles is crucial for maximum performance. In the future, when commercial vehicles become fully autonomous, this information needs to be accurate in order to guarantee optimal performance as well as safety. Accuracy is specifically important in components that are essential for the vehicle operation, such as the gearbox. Sensor solutions in modern gearboxes require an improvement, especially when it comes to low speed detection. This is important in automatic transmission, which is the preferred choice in commercial vehicles, so that the gearbox control unit can shift gears in optimal way.
This thesis project investigates the possibility of high accuracy speed and angular position detection in gearbox by multiple sensing. This was done with the use of distributed sensors covering the region of a single period on a gear structure. Hall sensors with analogue output were implemented for this purpose. Also, GMR sensors were introduced in this application, only in single unit measurement configuration.
The results show that the accuracy of low speed detection is higher with multiple sensors. Particularly in the case of braking down to zero speed, the accuracy is significantly increased in comparison to digital speed detection. However, at higher speed the accuracy is nearly the same. Therefore, one possible solution could comprise the use of multiple sensing at low speed and digital detection at high speed.
ISSN: 1654-7616, UPTEC E 17 007 Examinator: Mikael Bergkvist Ämnesgranskare: Uwe Zimmermann
Handledare: Andrey Gromov & Joakim Sommansson
Table of Contents
ABBREVIATIONS ... 1
1 INTRODUCTION ... 2
1.1 BACKGROUND ... 2
1.2 PROJECT DESCRIPTION ... 3
1.3 LIMITATIONS ... 4
2 THEORY ... 5
2.1 GEARBOX ... 5
2.2 SENSORS ... 6
2.2.1 Inductive Sensor ... 6
2.2.2 Hall Effect Sensor ... 7
2.2.3 Giant Magnetoresistance (GMR) ... 9
2.2.4 Sensor Technologies Comparison ... 11
2.3 QUADRATURE SIGNALS ... 12
3 IMPLEMENTATION ... 13
3.1 COMPONENTS ... 13
3.1.1 Selection of Sensors ... 13
3.1.2 PCB Design ... 15
3.1.3 Rig Assembly ... 17
3.1.4 Magnets ... 19
3.2 TESTING ... 20
3.2.1 Experimental Setup ... 20
3.2.2 Test Setup and Measurements ... 25
3.2.2.1 Single Unit ... 25
3.2.2.2 Array ... 27
3.3 PROBLEMS ENCOUNTERED ... 33
4 RESULTS AND DISCUSSION ... 34
4.1 SINGLE UNIT ... 34
4.1.1 Hall Effect ... 34
4.1.2 GMR ... 35
4.2 ARRAY ... 37
4.2.1 Four Hall Sensors ... 37
4.2.1.1 Separate Units – Sensor H ... 37
4.2.1.2 Integrated Units – Sensor K ... 46
4.2.2 Eight Hall Sensors ... 47
4.3 DISCUSSION ... 49
4.3.1 GMR ... 49
4.3.2 Multiple Sensing ... 50
4.3.3 Future work ... 51
5 CONCLUSION ... 53
REFERENCES ... 54
Abbreviations
DAQ – Data Acquisition DC – Direct Current
G – gauss: unit of measurement of magnetic flux density B GMR – Giant Magnetoresistance
Oe – oerstedt: unit of measurement of magnetic field H PCB – Printed Circuit Board
RPM – Revolution Per Minute SNR – Signal-to-Noise Ratio
V – volt: unit of measurement of electric potential (voltage)
1 Introduction
The demands placed on modern gearbox solutions to provide maximum performance and fuel economy coupled with the drive to reduce emissions, requires an increasing accuracy and specific functionality of sensors to monitor and help manage gearbox operation.
Despite the large variety of proposed sensors, it is not always possible to fulfil all
requirements to very specific sensor functionality suited to new gearbox design. Standard automotive manufacturers will insist on their own more general solutions that cover wider market.
Thus, in some areas, own prototypes or setup experiment series need to be created to investigate in details which sensor could be designed and manufactured according to
specifications on a custom basis. Present work deals with speed and angle detection devices that in different modifications should suite gearbox.
1.1 Background
As commercial vehicle systems become more complex and the automotive manufacturers strive to automate these systems in order to reduce the dependence of the vehicle’s operation on the driver, the information from numerous sensors in vehicles is essential. Eventually, the vehicles will become fully autonomous and it is important that this information is accurate.
Otherwise, safety and efficiency is not guaranteed and the overall performance of the vehicles cannot be improved.
One integral part of a vehicle that shall provide such information is the gearbox. Today, all vehicles with a combustion engine require a gearbox in order to operate. It is used for converting and transferring the torque from the engine to the drive wheels. A more detailed explanation about its function is found in section 2.1. The gearbox includes a number of sensors that measure the rotational speed and direction of the gears and shafts it comprises.
The output from the measurements is then used for providing information to the control unit that is responsible for gear changes. It needs to know when to couple or decouple the clutch or when to engage certain gears and shafts and synchronize their speed. All this is done at certain speeds i.e. revolutions per minute (RPM). Therefore, it is essential to apply sensors with high accuracy and measurement range in order to register speeds from 0 RPM to over 4000 RPM.
For this purpose, magnetic sensors have shown great reliability. They measure magnetic field changes during the gear rotation when teeth pass by the sensor. This requires a magnet to be placed behind the sensor. The magnetic field is produced because the gears are made of steel or iron-based material and the magnet is attracted more or less depending on whether a tooth is in front of the sensor or the space between two teeth.
There are a many magnetic sensors on the market today that differ in sensitivity, measurement and temperature range, output type, life span, cost, etc. Automotive manufacturers need to balance between these properties so they can get maximum performance for as low cost as possible. Sometimes, that is not the optimal solution and the overall performance and efficiency of the vehicle is compromised. Until now, most automotive producers have been using inductive proximity sensors in their gearbox. They are robust and reliable but they have a very low output signal at low RPM and they are
relatively expensive to produce. That is why there is a tendency for the next generation gearbox, which starts developing and will go into production soon, to use Hall effect sensors.
They have the ability to measure extremely low speed, they are smaller and cheaper and have high durability. However, in this particular situation, the chosen Hall have a digital output that leads to discrete measurements. Such discretisation, which is due to lack of continuous
information, results in low accuracy at very low speed. This work is dedicated to various solutions for continuous speed or angle measurements and will be implemented in the nearest future. Possible improvements will be investigated in this project that will present both experimental and theoretical analysis.
1.2 Project Description
The main objective of the project is to investigate and design next generation sensor for speed and angle detection in truck gearbox. The approach to the problem is to use distributed sensor elements for covering the region of single period of a pulse wheel or a gear structure. Then, as the gear wheel rotates, a measurement will be done continuously throughout the whole tooth period and there will be a signal output between the teeth as well. This requires sensors with analogue output in order to obtain the real speed of the rotation. The choice falls on sensor elements that measure magnetic field directly, independently of the speed. At the same time, it must be limited in size to maximum 1 mm. For this matter, Hall effect and GMR sensors that are already available on the market will be analysed and compared. GMR (Giant
Magnetoresistance) sensors are magnetic sensors with very high sensitivity that are becoming more popular because in some areas, such as computer industry, and in theory, they have shown better performance than Hall sensors. However, its maturity needs to be proven in practice, especially in automotive applications with harsh environment.
The first step in the project is to investigate the market for different types of mentioned sensors for automotive applications. Additionally, new prototypes and concepts on the market will be examined for further development. Next, each sensor will be tested separately in a rig assembly designed for testing speed sensors. This requires an adaptable experimental setup due to different sensor size and measurement technique. The experiment setup and the measurements will provide the basis for analysis.
Afterwards, selected sensors will be arranged in an array of several units. The arrays are also going to be tested in the rig and analysed accordingly.
Finally, a conclusion on which combination gives the best result for the purpose is to be given.
The measurements will be compared with the current gearbox sensor solution. Both electrical and mechanical requirements will be set and the technical functionality of the sensor will be described. The final result will include a description of output signal processing and different approaches and a proposed solution on the control unit requirements.
1.3 Limitations
Due to limited time for the project, certain scope limitations need to be set. As mentioned, only Hall and GMR sensors will be analysed although there are modern inductive sensor solutions that produce high accuracy measurements. The rig assembly used for testing can implement different types of gear wheels but in this case only three are used. The rig is also placed in a lab with different noise-generating equipment that cannot be displaced or shielded.
Finally, there is no possibility to perform a test in a real gearbox in order to see how the environment inside affects the output signal because of the preliminary stage of prototype.
2 Theory
This chapter explains the theory basics for better understanding of the implemented
technology. Primarily, a short explanation on the function of a gearbox is given and why it is important that they include high accuracy sensors. There are many different sensor
technologies that can be used in this application and the ones that are mainly implemented are described here. Finally, a brief comparison of the sensors is given.
2.1 Gearbox
As mentioned earlier, the gearbox is an essential component in vehicles. Its purpose is to convert the engine torque and transfer it to the wheels using different gears. There are many different types of gearboxes but the most basic are the ones with manual and automatic transmission. In a manual gearbox, the driver selects gears using a lever whereas in an automatic gearbox an actuator performs this operation. The gearbox is connected to the engine with a clutch or hydraulic coupling and the torque transfer to the wheel differentials is performed via a driveshaft.
The gearbox consists of different shafts with gears on. The input shaft is directly connected to the clutch, which in turn is connected to the engine. The clutch transfers the torque from the engine. In order to convert this torque, a layshaft is connected to the input shaft and it has different sets of gears on it. These gears mesh with gears on the main shaft depending on which gear is selected. The gear ratio determines the speed of the main shaft, which rotates with the same speed as the drive wheels. This is how the power from the engine is transmitted and converted through the gearbox to the wheels.
Gearboxes are constantly developed and they comprise many different parts that are
connected and rotating. So, in order for the transmission control unit to know when to change a gear i.e. which gears to engage and disengage, and when to couple or decouple the clutch, rotational speed of the shafts must be known. That is why it is essential to implement accurate sensing devices that can measure zero speed. This is important in automatic gearboxes
particularly, which are more common today in commercial vehicles, since the gears must be changed automatically at certain speeds. Additionally, modern trucks are made to operate at low torque because of fuel economy and this is another reason why sensors must be more precise and accurate.
Figure 2.1 below presents a possible sensor placement in a gearbox.
Figure 2.1: Possible sensor placement in a gearbox: 1) axial; 2) radial; [19]
2.2 Sensors
Different types of sensors can be used for speed detection in the gearbox. Today, inductive sensors are widely used in trucks. For the next generation gearbox, Hall sensors will be implemented since they have no speed limitation and are cheaper. In the future, GMR might be the primary choice in this application since it has higher sensitivity and it is more flexible.
These three sensor technologies are introduced and described below. Their main application areas are included as well as their advantages and disadvantages. Finally, a comparison between today’s choice, the Hall sensor, and possibly tomorrow’s choice, the GMR sensor, is presented.
2.2.1 Inductive Sensor
Detection of position and speed in automotive applications is most commonly performed with inductive sensors, partly because of their robustness and ability to function properly in severe conditions but also due to their reliability. Figure 2.2 presents the working principle of an inductive sensor for automotive applications. The sensor uses the principle of Faraday’s law of induction, which states that a voltage, e, is induced when the magnetic flux, Φ, in a coil with N turns is varying with time. This is described by the following equation:
! = !!Φ
!" (2.1)
The inductive sensor in figure 2.2 consists of a permanent magnet placed behind an induction coil with a soft iron core. Whenever an object of ferromagnetic material passes by the sensor, the magnetic field change, which implies a change in the magnetic flux through the coil. A voltage is then induced in the coil and present at the output. [1]
Besides the automotive industry, inductive sensors are used in areas where high reliability is crucial. These include heavy industry, robotics, military, aerospace, etc. [2,3]
As mentioned, inductive sensors are very reliable in extreme conditions and this is because of their robustness and contactless operation. They can also operate in high temperature, above 200°C, and because of all of this, they are long-life sensors. However, their robustness is a disadvantage too since they are relatively big and heavy and cannot physically fit into many places in certain applications. They are also expensive to produce because of their design and material requirement. [2,3]
Figure 2.2: Inductive sensor in automotive applications [1]
2.2.2 Hall Effect Sensor
Hall effect sensors are magnetic sensors, which implies that they respond to external magnetic field. Their working principle is based on the Hall effect phenomenon that occurs when a magnetic field is present in the proximity of an electrical conductor and is perpendicular to the current. A voltage difference is then produced across the conductor and it is called the Hall voltage. The Hall effect was discovered in 1879 by American physicist Edwin Herbert Hall. The working principle is presented in the left illustration of figure 2.3 below.
In Hall sensors, the Hall effect occurs in a doped semiconductor that the sensors consist of.
When the semiconductor is fed with voltage, often 5 V, a constant current flow is created.
With an external magnetic field applied, the charge carriers get deflected and electrons and holes are placed on the either side of the semiconductor plate. This creates an electric potential difference i.e. a voltage difference, which is measured. This is the mentioned Hall voltage and it is directly proportional to the magnetic field strength. The voltage lies often in the mV range and is amplified with an internal DC-amplifier.
The following equation defines the Hall voltage, VH:
!! = !! !
!×! (2.2) Here, RH is the Hall coefficient that characterizes the conductor material, which usually
reflects the electro mobility, I is the current flow, t is the sensor thickness and B is the magnetic field.
Hall sensors can have both linear i.e. analogue and digital output signal. The linear output is the raw voltage signal that is directly proportional to the magnetic field and digital output is presented with states “ON” and “OFF”.
One Hall sensor may consist of one or several Hall elements integrated on the chip. With one element, a direct measurement of the magnetic field is possible. With two elements, a
differential output is available. A Hall sensors consisting of more than two Hall elements can be used for more complex and accurate measurements by measuring the phase shift between the differential signals from each sensor pair.
Since magnets have two poles, north pole and south pole, there are Hall sensors that require either one or both poles for operation. Sensors that only need south pole i.e. a positive
magnetic field, to have an output signal and to switch from LOW to HIGH and vice versa, are called unipolar. Hall sensors with an output that switches from LOW to HIGH when a
positive magnetic field is present and back to LOW when a negative magnetic field i.e. north pole is present instead, are called bipolar. Hall sensors can also operate with a back-biased magnet, which is a magnet placed behind the sensor, and detection of ferromagnetic materials placed in front of the sensor is possible in this manner [4-6]. This is presented in the right illustration in figure 2.3.
Hall sensors are widely used in low-power applications in the automotive industry and industrial electronics. Some examples are position and speed detection, current sensing, contactless switching, computer keyboards, etc. [6].
Figure 2.3: left: Hall sensor working principle [4]; right: Hall sensor with back-biased magnet [20]
The many advantages of Hall sensors are the reason why they are the most commonly used sensors for magnetic field detection. Physically, they can be very small, making it possible to apply them in places where space is critical, and harsh environment does not affect them.
Regarding the operational aspects, Hall sensors can manage frequencies up to 100 kHz, they can measure zero speed and they have a relatively wide temperature range [7,8]. Also, Hall sensors are very cheap to produce which puts them “among the most cost-effective magnetic field sensors available today” [7].
However, there are a couple of disadvantages that these sensors have. They can be subject to interference from external magnetic fields other than the field that is supposed to be measured.
There is also a limit to the distance between the sensor and the measurement object depending on how strong magnet is applied. [8] Some other negative sides are the fact that Hall sensors are sensitive to temperature drift, which requires compensation, and that they always have an offset voltage due to non-ideal fabrication [6].
2.2.3 Giant Magnetoresistance (GMR)
GMR was discovered in the late 1980s and since then it has been the primary sensing technology in various computer applications, mainly for hard drives and random-access memory, as well as in life science where magnetic biosensors are used [9]. GMR sensors have also been applied in navigation systems, current sensing, angle measurement, etc. [10].
This nanotechnology works on the principle of layering thin magnetic and non-magnetic films.
The simplest form of doing so is by using the spin valve arrangement that comprises two ferromagnetic layers separated by a non-ferromagnetic conductive thin layer. One of the ferromagnetic layers has a fixed magnetization and therefore it is usually called the pinned layer or the hard layer. The magnetization in the other magnetic layer, the free or the soft layer, is free to rotate because of the antiferromagnetic coupling. When the magnetization direction in the two layers is opposite, the resistance to the current flow through the non- ferromagnetic layer is high. If an external magnetic field is applied parallel to the arrangement, the magnetization in the free layer rotates and it becomes oriented parallel to the pinned layer.
This lowers the current resistance. The change in resistance can be measured and that is how the presence of external magnetic field can be sensed. [11, 12] This change can be up to 20%, which is a lot in comparison to other technologies and that is why it is called giant.
The working principle of GMR sensors is presented in figure 2.4.
Figure 2.4: GMR working principle [21]
In addition to the basic spin valve arrangement, GMR sensors are also built with multiple layers in order to increase the sensitivity of the device. Typically, a GMR sensor element is a Wheatstone bridge with two sensing resistors and two reference resistors. The differential measurement gives an output voltage that is proportional to the applied magnetic field [11].
Figure 2.5 presents a typical output and Figure 2.6 presents a Wheatstone bridge.
Figure 2.5: Typical GMR sensor output [13]
Figure 2.6: Wheatstone bridge with differential output. Source: Illustration by author As mentioned earlier, GMR sensors are used in hard disk drives for reading data. Also, the spin valve configuration is used in magnetoresistive random-access memory (MRAM) where the direction of the magnetization in layers is used for encoding the stored bits. Recently, the automotive industry has started to implement GMR sensors in various applications, such as speed, position and angle measurements. This is because of their high sensitivity, small size, low power consumption, low noise and temperature stability. Some applications include the crankshaft position and speed, the wheel speed, the steering angle [14].
2.2.4 Sensor Technologies Comparison
Depending on the application, the described sensor technologies have their advantages and drawbacks. In the automotive industry, inductive sensors and Hall effect sensors are
predominant due to their reliability. Although inductive sensors are being replaced because of their robustness and high cost, they still remain the primary choice in applications where high temperature and rough conditions are present. In speed sensing, they are replaced by Hall sensors since inductive sensors have a measurement limitation at low speeds. Therefore, the best choice for gearbox speed sensing is Hall effect and GMR sensors. Both of these are cheap to produce, they have high sensitivity and can measure zero speed. GMR sensors have an overall better performance than Hall sensors and theoretically they are advantageous regarding the operational aspects. Because of the fact that they are sensitive to a magnetic field parallel to the sensor chip, they are more flexible and can perform the same
measurement in different practical setups. Since they are more sensitive, there can be a larger air-gap between the sensor and the measured object, creating more degrees of freedom. All this is theoretically accurate but in order to prove a clear advantage of GMR sensors relative to Hall sensors, a GMR sensor with 10-100 times higher sensitivity needs to be built. That is physically impossible with today’s technology but it has been presented in theory. Also, for automotive applications, there are no series produced GMR sensors ready for use, only on- chip solutions. Due to this fact, Hall sensors are preferred and will remain so for some time.
Table 1 below presents electrical characteristics and differences between Hall and GMR sensors.
Table 1: Characteristics of Hall effect and GMR technologies [15]
Hall Effect GMR Power consumption (mA) 5 - 20 1 - 10 Field sensitivity (mV/V/Oe) ~ 0.05 ~ 3
Dynamic range (Oe) ~ 10000 ~ 100
Temperature performance
(°C) < 150 < 150
Temperature stability low high
Die size (mm2) 1 x 1 1 x 2
Signal level small large
Cost low low
2.3 Quadrature Signals
In signal processing, it is quite common to implement quadrature signals. These are periodic and phase shifted with 90°. Most often, the term quadrature is used in case of sine waves.
When, for example, four sinusoidal signals with a 90° phase shift are to be processed, they would be paired up in order to obtain quadrature signals, i.e. two sinusoidal signals. This is explained with the expressions below. The four sinusoidal signals, y1, y2, y3 and y4 are defined as:
!! = !"#(!") (2.3)
!! = !"# !" + 90° (2.4)
!! = !"#(!" + 180°) (2.5)
!! = !"#(!" + 270°) (2.6)
Here, !" denotes the angle, where ! is the angular frequency and ! is the time. In order to obtain quadrature signals, signals with a 180° phase shift are paired up. In this case, y1 is
paired up with y3 and y2 is paired up with y4. These are subtracted and the quadrature signals,
! and ! are derived:
! = !!− !! = !"# !" − !"# !" + 180° = !"#$%&%'(!"#) !"#$%!%!#&
= 2sin (!") (2.7)
! = !!− !! = !"# !" + 90° − !"# !" + 270° = !"#$%&%'(!"#) !"#$%!%!#&
= 2cos (!") (2.8)
Thus, signals ! and ! are 90° phase shifted. In order to obtain the angle !", the inverse tangent is calculated:
!"# !" =!"# !"
!"# !" =!
!
(2.9)
⟺ !" = !"#!!(!
!)
This method can be implemented when processing the signals from sensors arranged in an array.
3 Implementation
This chapter provides a detailed description of the methodology and the practical work implemented in this project. The practical part is the main section here since there were many components and experimental setups included in the working process.
First, the sensors that were selected for the purpose are described with regards to their technology and main characteristics. A table with the characteristics for each of them is presented for better understanding and comparison. The sensors were tested separately at first and then arrays of selected units were set up. For this purpose, a PCB was designed and printed accordingly, which is presented below. Also, the rig assembly that was used for performing the tests is shown.
Next, the experimental setup is described. In order to perform the experiments properly, some equipment needed to be adjusted and certain considerations, such as circuit board size, wiring, placement of sensors, etc., had to be made. Different types of measurements that were
performed, both with single units and arrays, are presented.
Finally, all the problems that were encountered during the tests are given.
3.1 Components
3.1.1 Selection of Sensors
As mentioned in the project description, only Hall and GMR sensors were considered for the purpose. There is a need for a better variant of the Hall sensor than the one that will be used in the new gearbox but further market investigation on GMR technology for automotive
applications must be done as well. The current sensor has a digital output signal that only registers a gear tooth passing by and it gives no information on the speed between the teeth.
This discrete measurement method fails particularly in situations when the gear slows down and the time period between the teeth is longer. In that case, the error in speed detection is much larger since the output from the sensor shows the speed that was registered at the last passing tooth. Also, if the rotation is stopped when the sensor is between two teeth, the actual zero speed will be displayed as the recorded speed from the last tooth. The signal goes down to zero eventually but this could be an great issue in high technology vehicles, such as autonomous vehicles, in the future.
The next generation sensor would need to have an analogue output signal that would give information continuously, which would then be processed externally for optimal function.
This was the main consideration during the selection process. The requirement was to find sensors with raw signal available at the output and no integrated circuitry that processes this signal (except an internal amplifier possibly, which could be necessary if the signal is too low). Another aspect that was taken into account was the size of the actual chip. This was important for the testing of arrays, since an array needed to cover one tooth period on a gear wheel and its length could not exceed the length of the period. In this context, it was also better to choose sensors chips with no outer pins that extend the overall size of the component.
By investigating the market, it appeared that all sensors had some similar electrical ratings, such as input voltage, maximum input current, resistance, maximum temperature. One
essential characteristic that differed between the components was the sensitivity. Especially regarding GMR sensors, there was a wide range of magnetic field sensitivity, which implies more flexibility when it comes to the choice of magnets, air-gap, etc.
The final selection of sensors is presented in Table 2. Various features and characteristics are included, with some information missing for certain units. The elements are shown in figure 3.1. Note that Sensor K is a component with 4 integrated Hall elements and has both analogue and digital output. This component was used as an array only.
Table 2. Selected Hall and GMR sensors with characteristics
Sensor
label Technology Type Output Sensitivity Saturation
Supply voltage [V]
Temp.
Max.
[°C]
Size [mm]
Min Max Min Max
mV/V-Oe Oe/5 kΩ
A
GMR
Field
Single bridge analogue
11.0 18.0 6 < 1 ± 12 +150
5 x 4
B Low
Hysteresis 3.0 4.2 15 < 1 ± 25 +150
C Field 0.9 1.3 50 < 1 24 +125
D Field 0.45 0.65 100 < 1 24 +125
E Gradient - - 250 < 1 12.5 +125
F Low
Hysteresis - - ± 180 Oe < 1 30 +150
3 x 3
G Low
Hysteresis
Dual bridge
analogue - - ± 180 Oe < 1 30 +150
H
Hall Field
Ratiometric
analogue 3.125 ±
0.094 mV/G - 4.5 10.5 +125 4 x 3
I Ratiometric
analogue
10
mV/G - 2.5 3.5 +85 2 x 3
J Analogue &
Digital - - 4.5 36 +125 4 x 4
K Analogue &
Digital -0.3 6 +125 6.2 x 4.2
Figure 3.1: The sensors that were used in the project
3.1.2 PCB Design
In order to construct arrays of sensors, a PCB was designed and printed for some of the units.
It was important to have the circuit board as thin as possible so that a back-biased magnet could be placed close behind the sensor. Also, the size of the board had to be appropriate for flexibility purposes during the test setup and execution.
The PCB was designed in Target 3001! [16], and figure 3.2 presents a drawing of it.
Figure 3.2: PCB made in Target 3001!, top: drawing, bottom: final design
The top left array circuit board was printed according to the drawing, not as presented in the final design picture in the figure above. This was done for better convenience.
It must be mentioned that the PCB was designed before the final selection of sensors was made. This resulted in a PCB suitable for only four sensor types. Namely, the four boards to the left (see figure 3.2) are for sensor J only. Two boards in the upper-right corner of the PCB are for sensor I. In the lower-right corner, the two boards are for sensor F and G respectively.
The arrays were designed so that the elements could be placed as close to each other as possible in order to limit the length of the arrays but also to comprise as many elements as possible for improved measurement.
As it can be observed in figure 3.2, different types of arrays were designed. The simplest ones consist of a single row with selected number of elements. The other ones are slightly distinct.
The top left one contains eight elements of sensor J but due to its size, two rows of four elements were made instead. The rows are shifted with one half-length of the sensor. The bottom left was supposed to be an arc with five elements but due to wrong radius, it was not implemented. Instead, an arc with a radius that corresponds to the gear wheel with long tooth period, see section 3.1.3, was designed. This is the board placed in the middle and it can contain five elements.
The sensors were placed on all circuit boards described above and the wiring was completed but due to time limitation, only one of these was actually tested, namely the array with sensor I.
The final PCB is shown in figure 3.3.
Figure 3.3: PCB printed out
3.1.3 Rig Assembly
All experiments were performed in a test rig that was constructed specifically for testing different speed sensors on different sets of gears. The design was made to resemble a gearbox shaft and its operation. Figure 3.4 presents the rig assembly.
Figure 3.4: The test rig
The rig is operated using a LabVIEW interface, an Arduino board, a DC motor and an
electromechanical brake. Via communication between the interface and the Arduino, the shaft in the rig can be rotated. In the interface window, one can connect the motor to power, set the rotational speed to over 1000 RPM and stop the rotation either by using the brake or a slow stop function that gradually slows the rotation down to 0 RPM. All settings and parameters in the LabVIEW interface are sent to Arduino, which controls the motor and the brake by sending out the right signals to respective component. A DAQ device for monitoring the output data when testing a sensor is also included in the setup but it malfunctioned in this project so an oscilloscope was used instead. A detailed description of the mechanical construction and the software of the assembly is found in [16] and [17] respectively.
The shaft that rotates is installed in the manner that a gear can be mounted on one end of it for easier approach. That way, different sensor measurements can be performed. Gear wheels of different type and size, which are used in the gearbox, can be installed. In this project, mainly two wheels were used in the experiments. Figure 3.5 shows both of these.
Figure 3.5: Left: two wheels, inner and outer, used for sensor testing;
Right: Teeth of the outer wheel
The wheel with the smaller radius originally included additional wider teeth that were placed between the small teeth shown in the picture. These were removed in order to obtain equally distributed teeth with small width and large distance between them. This type of wheel was primarily used in axial measurements of single unit sensors just to register what kind of output the sensors had when a tooth passed by. It was used in some array measurement as well in order to verify that all sensors in the array have the same output and react similarly during the rotation.
The other wheel applied in the measurements has the right-hand helical teeth arrangement and it corresponds to a type of gear that is actually applied in the gearbox. That is why the radial measurements that were performed on it are essential and provide the most important and usable results. It was used for all array measurements and some GMR single unit
measurements.
At the final stage of the project, another wheel was implemented for one array measurement.
It is the flywheel shown in figure 3.6. It was used mostly due to the similarity in radial measurement as above but in this case the wheel had straight teeth instead of helical.
Section 3.2 describes which wheel was applied in each measurement.
Figure 3.6: The flywheel used in one array measurement
3.1.4 Magnets
Two different magnets were used for back-biasing the sensors during the tests. The magnets’
size was considered so they could cover a whole sensor chip or one whole array of sensors.
Unfortunately, information about the strength of the magnets was not presented in their respective datasheet but in practice the difference in strength was evident. The magnets with the magnetization direction are presented in figure 3.7.
Figure 3.7: Magnets used for back-biasing the sensors
The magnet shown in the top picture turned out to be much stronger than the other magnet. It was so strong that in one situation, it managed to demagnetize the weaker one a bit when they were in contact. That is why it was not used as often in the experiments since the sensors output reached the saturation level when the magnet was placed close. It was only applied in some situations for comparison purposes and in certain setups where it was practically more suitable to use instead of the weak magnet. Also, the weaker magnet was better to use since it has its poles on the short sides, which are square and can back-bias each sensor almost
perfectly.
3.2 Testing
3.2.1 Experimental Setup
The conducted experiments required certain equipment to be applied. Some adjustments and improvised solutions, especially regarding the placements of circuit boards, were
implemented for easier performance.
For each test, a power supply unit was used to power the sensors and the output signals were displayed on an oscilloscope. As mentioned in section 3.1.3, the rig was operated from a computer. Also, a multimeter was used for troubleshooting and verification purposes.
The circuit boards had to be placed correctly relative to the gear wheel in order to obtain optimal output signal. More precisely, the sensors had to be centred and positioned as close as possible to the gear wheel. This was possible with a special placement station shown in figure 3.8.
Figure 3.8: A placement station for the sensors
The station has a millimetre scale that is used for placement adjustments in all directions. This was very useful for choosing the air-gap between the sensor and the gear. Since the whole rig was sitting on a thick metal plate, the station could be fixed to it with magnets that are placed underneath it and are shifted down for attachment or back up for detachment using the two knobs seen in the picture.
This station required an additional part attached on top that was carrying the circuit board and the magnet and which could reach the same height as the centre of the gear, where the
measurements are performed. This carrier piece is presented in figure 3.9.
Figure 3.9: A carrier piece for sensor and magnet placement
An opening was made for the weaker magnet to be placed behind the sensor. This way, the magnet could be as close to the sensor as possible with only the circuit board separating them.
This type of arrangement was used only in single unit measurements since the arrays required several magnets that could generate a magnetic field covering the whole array. In that case, four magnets of the weak type were put together with the same polarity on the ends, see figure 3.10.
Figure 3.10: Magnet arrangement for back-biasing in array measurements
An example of the whole experimental setup is presented in figure 3.11. In this case, an array measurement is shown. It can be seen how the placement station and the carrier piece are attached together.
Figure 3.11: The experimental setup in array measurement configuration
All single unit sensors and some arrays were soldered by hand. The rest was done in a reflow oven. Figure 3.12 shows an example of a single unit sensor soldered on a circuit board.
Figure 3.12: A GMR sensor (sensor E) placed on a circuit board with wiring
The other elements were done in the same manner with consideration of choosing a circuit board as thin as possible so there would be minimum distance between the sensor and the magnet.
The sensors that were tested in the array configuration are sensor H and I as well as sensor K, which has an integrated array of four elements. The Results section below describes all the choices made regarding the array measurements.
The array consisting of sensor H elements had to be designed on a regular breadboard since sensor H has straight leads (se figure 3.1) and an adequate PCB cannot be designed for it, unlike for flat no-lead packages. Figure 3.13 presents the array which consists of six elements.
Figure 3.13: Array with six sensor H elements
The arrangement consists of two rows with three elements each and a half-length shift. Two separate breadboard pieces were used so that a magnet could be placed behind the array. The magnet setup is shown in figure 3.14.
Figure 3.14: Placement of magnets behind the sensor H array (top view)
An additional unit was added to one of the rows in order to create a four-element array for another measurement, see figure 3.15. The same magnet solution was implemented in this case.
Figure 3.15: Sensor H array with four elements
The PCB that was designed for the purpose was used in sensor I array configuration of eight units, presented in figure 3.16. This sensor is the smallest size and that is why it was possible to include so many units on one straight-line array. The back-biasing with magnets was done the same way as above.
Figure 3.16: Array configuration of sensor I
3.2.2 Test Setup and Measurements
This section provides a description of the measurement types performed in the experiment. It is presented what settings were used, which gear was used for each measurement and what adjustments were applied practically. A number of figures are presented for better
understanding.
The section is divided into two subsections that include single unit measurements and array measurements respectively.
3.2.2.1 Single Unit
Single unit measurements were done similarly for each sensor type. The input voltage was set to 5 V in each case except for sensor I, which has a lower input voltage range and it was supplied with 3 V instead. Since single unit measurements were performed mostly for the purpose of inspecting what kind of output signal the sensors had when a gear tooth passed by, mainly axial measurements were done. Axial measurements are performed on the smaller gear which has a low number of narrow teeth and a large distance between them (see figure 3.5).
The measurement is called axial since it is performed in parallel with the rotation axis unlike radial measurements, which are done in parallel with the radius of the gear. Radial
measurements were performed on the gear wheel that has helical teeth arrangement. Table 3.1 shows which type was implemented for each sensor.
Table 3.1: Type of measurement performed for each sensor
Sensor A B C D E F G H I J
Axial yes yes yes yes yes yes yes yes yes yes
Radial no no yes no no no yes no no no
The radial measurement was performed with GMR sensors since GMR is sensitive to planar magnetic field, which implies a different output signal form with different arrangements. This will be presented in the Results section.
Each sensor was tested at a rotational speed of 50 RPM and 500 RPM. The higher speed was included only for comparison purposes, namely to verify that the output signal form was not affected. At the beginning of each test, the sensors were placed at the minimum possible distance to the gear wheel without physical contact and measured. Then, for the following measurements, the air-gap was increased with 1mm in each case using the millimetre scale on the placement station. This was done in order to see how sensitive the sensors were and what the maximum allowed air-gap was.
All sensors were tested with the weaker magnet placed behind them at an estimated distance of 2-3 mm due to the circuit boards. Only sensor H was measured with the stronger magnet too but at a larger distance. This was done just to compare the output when using magnets of different strength.
When all measurements were finalized, the raw data was analysed with respect to the output signal characteristics such as strength, signal-to-noise ratio, offset, sensitivity, smoothness.
Figure 3.17 below shows the sensor placement during an axial measurement. All single unit axial measurements were performed using the presented arrangement.
Figure 3.17: Axial measurement
Figure 3.18 presents the configuration of a radial measurement, which was performed with GMR sensors and generated important results.
Figure 3.18: Radial measurement
3.2.2.2 Array
Four different arrays were tested in this project, namely arrays with sensor H (see figure 3.13 and 3.15), sensor I (see figure 3.16), and sensor K. Like single units, these were also
implemented in axial and radial measurement. Table 3.2 presents what measurements were done for each array, how many sensors were included and in what arrangement, and what the sensor element pitch was.
Table 3.2: Array measurement and setup characteristics
Sensor Elements Arrangement Pitch (mm) Measurement Axial Radial Flywheel
Radial
H 4 Single row 4 no yes yes
H 6
Double row with half-length shift,
3 elements/row 2 yes no no
I 8 Single row 2 yes yes no
K 4
(integrated) Single row 1.28 no yes no
First, the array with six units presented in figure 3.13, section 3.2.1, was tested in axial measurement in order to see how the signals would be shifted and how much offset each sensor would have. The setup configuration is shown in figure 3.19.
Figure 3.19: Test setup for sensor H six-element array in axial measurement
Due to this type of arrangement, with two rows of sensor arrays, it was not suitable to perform a radial measurement. Instead, one sensor was added to one row in order to obtain a four- element single-row array, see figure 3.15. This was then used to perform several tests on the wheel with helical teeth as well as one test on the flywheel. Figure 3.20 presents the first case.
Figure 20: Test setup for sensor H four-element array in radial measurement As it can be seen in the picture, the array had to be tilted a few degrees in order to cover one whole tooth period. It was done so because the sensor pitch was not ideal for the wheel used in the test. Ideally, with the right pitch, the array would be placed straight across one tooth period, as illustrated in figure 3.21.
Figure 3.21: Ideal horizontal sensor array placement over one (helical) tooth period Note: left-hand helical gear used in illustration
The first test was performed at a constant speed of 50 RPM and 500 RPM. In the second test, the signal output was to be analysed when braking the wheel. Braking was done from 25, 50 and 500 RPM with the electromechanical brake. Also, the slow-stop function in the program was used, where the wheel gradually slowed down from 50 RPM to 0 RPM.
This arrangement was tested on the flywheel as well. Constant speed of 25 RPM and 500 RPM were measured. Then, instead of using the electromechanical brake, the rotation was stopped by disabling the motor i.e. by turning off the supply to the motor. The LabVIEW interface included this feature. This was done because the motor generated noise which was noticeable in the signal waveforms. When the motor was turned off, the rotation gradually stopped but a couple of signal periods without noise could be registered before it had stopped completely. This solution was applied in sensor I measurements as well.
The flywheel test setup is given in figure 3.22.
Figure 3.22: Test setup for sensor H four-element array in flywheel measurement In this case, a larger tilt had to be fixed. Also, the array could not be placed vertically straight across the teeth since its length exceeds the length of a tooth period.
The eight-element array using sensor I, see figure 3.16, was tested both axially and radially.
The solution of turning off the motor and registering the signals was implemented. However, higher rotational speed had to be set in order to capture one whole signal period without the noise coming from the motor. Precisely, 80 RPM and 120 RPM were applied. This was done for both axial and radial measurement. The two setups are presented in figure 3.23. Even in this case, in the radial measurement setup, the array was slightly tilted although this is difficult to notice in the figure.
Figure 3.23: Axial and radial measurements setup for the sensor I array
Sensor K was tested only on the wheel with helical teeth since the sensor pitch and the tooth period conformed almost perfectly. The sensor pitch of 1.28 mm resulted in a total pitch of 5.12 mm with four integrated sensors and the distance between two teeth on the gear was approximately the same. An exact number could not be given due to the lack of a suitable measurement device. In this case, the integrated array could be placed perpendicularly across the teeth as illustrated in figure 3.24.
Figure 3.24: Ideal perpendicular sensor array placement over one (helical) tooth period.
Note: left-hand helical gear used in illustration
Since sensor K had an integrated signal conditioning as well as additional circuitry for both analogue and digital operation modes, it was possible to set different configurations and obtain different types of output.
Figure 3.25 shows the block diagram of sensor K.
Figure 3.25: Sensor K block diagram
By configuring the inputs 1, 2 and 3, two modes were selected for the measurement purpose.
In the first mode, positive sine and cosine signal with their corresponding negative signals could be obtained by internal sensor pairing with 180° phase shift. Here, the reference level is set to half the input voltage. The second mode generates a sawtooth waveform that lies within two threshold voltages that are set externally.
Figure 3.26 shows how the output of these two modes would look like.
Figure 3.26: Ideal output of two different modes of sensor K
These two modes were implemented in the test. An electrical switch was used together with a voltage divider in order to switch between the modes. Table 3.3 gives the configurations settings for the modes.
Table 3.3: Sensor K:s two operation modes. D-Sensor and R-Sensor denote the name of the modes
The test was performed for constant speed, braking, slow stop and acceleration in both operation modes. In each case, the speed was set to 25, 50 and 500 RPM.
The test setup is presented in figure 3.27. Here, the stronger magnet was used since it was practically easier to apply. Because of its strength, it was placed approximately 2 cm behind the sensor, which can be observed in the figure below. The tilt of the sensor was fixed in the manner so that the sensor was parallel with the teeth i.e. so that the array was perpendicular to the teeth, as in figure 3.24.
Figure 3.27: Sensor K test setup
3.3 Problems Encountered
Due to the practical setups and adjustments that had to be made, a number of issues have occurred during the experiment. These caused imperfections in the measurements and delay in the working process. Therefore, some prepared tests were not performed.
First of all, soldering by hand was not done flawlessly due to the size of the sensors. Since they are very small, it was difficult to handle them. One problem was to place the same amount of soldering paste for each pin of the element. If minimal errors were made, the sensor would not sit flat on the circuit board. There would be a small inclination and although it would not be visible, it would affect the sensor output. The magnetic field would then be sensed at a different angle, causing a different output. The same issue would come up if the sensors were not placed perfectly in parallel with the gears or if the magnet was not placed perfectly sensor. Small deviations would then result in a slightly changed output. This was crucial especially for GMR sensors in axial measurements since they are sensitive to the planar magnetic field. If one side of the sensor was closer to the gear than the other sensor, then the magnetic field components would be of different size and direction from side to side.
Another soldering issue occurred when using the reflow oven for array configurations. Due to the size of the sensors, specifically sensor I, some elements could be misplaced during the operation in the oven because of the airflow inside. That would imply an imperfect alignment of the array elements and the output would differ. Some arrays were not tested because of this problem.
During the radial measurements using the arrays configurations, it was difficult to tilt the array in accordance to the helical teeth and to cover one tooth period accurately. Several measurements of one array would be performed with different tilt for comparison in order to find the best solution.
There were also some issues with the rig performance. All measurements were affected by noise from the surrounding equipment but the major source of noise was the DC motor. It was clearly noticeable on the oscilloscope, having a periodic waveform with constant amplitude.
In some tests, it was eliminated by applying the solution presented in 3.2.2.2. Also, when the brake was applied, the whole construction would vibrate and the transition from a certain speed to 0 RPM would be difficult to register in the signal waveforms due to high transients.
In some situations, the brake would also be blocked and it would take time to release it. This was probably an issue in the signal transmission from the Arduino to the relay in the brake.
4 Results and Discussion
This section presents the results obtained in the project. Firstly, single unit measurements with Hall and GMR sensor are given. Only representative results are shown i.e. certain sensors are omitted due to the similarity in output and behaviour. Secondly, the results from array
measurements are described. Even in this case, only presentable results are given. A number of plots are shown representing raw data and the processed data. The most useful raw data was processed in Matlab and compared with the current gearbox sensor solution.
Finally, the overall performance and outcome is discussed and future work is suggested.
4.1 Single unit
4.1.1 Hall Effect
The raw data presented here is obtained from sensor H and sensor I. Figure 4.1 and 4.2 show their respective output signals. It can be seen that the output voltage is directly proportional to the external magnetic field. The axial measurement configuration was used in both cases and the signals resemble the tooth arrangement of the wheel. The signals were measured at 50 RPM sensor H and 500 RPM for sensor I. Comparing these two, the shape of the waveform is the same but sensor I has a lower amplitude since it has a lower output range. As mentioned earlier, it was supplied with 3 V instead of 5 V, which was used for all the other sensors.
Both signals were filtered with a 1 kHz internal filter on the oscilloscope.
Figure 4.1: Sensor H single unit output voltage at 50 rpm
Figure 4.2: Sensor I single unit output voltage at 500rpm
4.1.2 GMR
GMR sensors were tested both axially and radially. Since all sensors have the similar output, only one sensor is considered here and it is the representative for the rest of the units.
The axial measurement at 50 RPM is presented in figure 4.3 and the radial measurement, also at 50 RPM, is presented in figure 4.4.
Figure 4.3: Sensor C axial measurement output voltage at 50 rpm
Figure 4.4: Sensor C radial measurement output voltage at 50 rpm
In figure 4.3, the signal is saturated due to high sensitivity of the sensor that was distanced only 1 mm from the wheel. The sharp drop occurred when the centre of a tooth passed by and the tops were detected at the tooth edges.
Figure 4.4 shows a regular sine signal, which was obtained when the sensor was placed on the edge of the helical gear.
The difference between these two signals is discussed in section 4.3.1.
4.2 Array
4.2.1 Four Hall Sensors
This section will present the results from two arrays containing four hall sensor elements, namely sensor H and sensor K with integrated elements. Numerous plots will be given with raw and processed data with a brief description of the signal processing method. The results will be discussed in section 4.3.2.
4.2.1.1 Separate Units – Sensor H
The results from radial measurement on the helical gear are presented. For each test, the raw data will be shown followed by plots with the processed data.
First, constant speed of 50 RPM is presented in figures 4.5, 4.6 and 4.7. The first plot gives the output voltage from each sensor. These signals were processed in Matlab, where offset compensation was done. Then, sensors with a physical shift of 1/2 period, or a 180° phase shift, were paired up and their corresponding signal were subtracted. This resulted in quadrature signals presented in figure 4.6. The plot in this figure shows also the calculated angle from these two signals, which was obtained by calculating the arctangent. This angle gives the information on how many degrees the wheel has rotated within a certain
measurement range, with the reference angle i.e. 0°, chosen at the start of the measurement. In figure 4.6, the angle for one tooth period is given, only with converted unit of measurement presenting the angle within a range of 0 V to 1 V. Figure 4.7 presents the same angle in degrees.
Figure 4.5: Output voltage from sensor H radial measurement at 50 RPM
Figure 4.6: Subtracted signals with sensor pairing and the calculated angle
Figure 4.7: The calculated angle shown in degrees. Speed: 50RPM
Second, the slow stop function is presented where the rotation gradually goes down from approximately 30 RPM to 0 RPM. The results are shown in figures 4.8, 4.9 and 4.10 and 4.11.
Same signal processing method was applied in this case but with an additional step. Namely, the speed was obtained by calculating the derivative of the angle. In figure 4.10, the angle is presented for the whole measurement range, not for one tooth period only. The green trace in figure 4.11 presents the raw speed, which was obtained by calculating the derivative at each point of the angle curve. The red trace is the averaged speed calculated using the least square method. The black step line shows the simulated speed of obtained form digital measurement.
Figure 4.8: Output voltage with slow stop from ~30 RPM
Figure 4.9: Subtracted signals with sensor pairing and the calculated angle at slow stop
Figure 4.10: The angle, the raw speed, the averaged speed and the simulated speed from digital detection at slow stop
Figure 4.11: The raw speed, the averaged speed and the simulated speed from digital detection zoomed in
Next, the signals obtained when the brake was applied at 25 RPM are presented. The raw signals and the processed signals can be observed in figures 4.12, 4.13 and 4.14.
Figure 4.12: The output voltage with applied brake at 25 RPM
Figure 4.13: Subtracted signals with sensor pairing and the calculated angle
Figure 4.14: The angle, the raw speed, the averaged speed and the simulated speed from digital detection
Figures 4.15, 4.16 and 4.17 present the signals obtained during the attempt to stop the rotation manually. The slow stop function was applied in this situation and when the rotation slowed down to a certain speed close to 0 RPM, the shaft was physically pressed in order to stop the wheel. The signals reflect how the power from the motor counteracted the physical power that was trying to stop the wheel.
Figure 4.15: The output voltage during the manual stop
Figure 4.16: Subtracted signals with sensor pairing and the calculated angle
Figure 4.17: The angle, the raw speed, the averaged speed and the simulated speed from digital detection
Figures 4.18, 4.19 and 4.20 present the signals from the acceleration of the wheel.
Figure 4.18: The output voltage during acceleration
Figure 4.19: Subtracted signals with sensor pairing and the calculated angle
Figure 4.20: The angle, the raw speed, the averaged speed and the simulated speed from digital detection
4.2.1.2 Integrated Units – Sensor K
In this section, the measurement results from sensor K, which has four integrated Hall elements, are presented. The test was performed at 50 RPM constant speed in both
configuration modes. Figure 4.21 shows the output using the mode with sinusoidal signals and figure 4.22 present the sawtooth mode with no filter.
Figure 4.21: Output voltage in the sinusoidal configuration mode at 50 RPM
Figure 4.22: The sawtooth output voltage at 50 RPM and with no filter applied
